Well, this comes to ~50-55 MBs xfer rate from the drive media to the cache on the drive... of course you can burst from the 2MB cache at up to bus speed (more or less), and if you are on a slower bus (plain old UW, etc) you could do a non-cached burst... that being said, if all you are doing is serving static files, then - yeah, the bottleneck isn't the drive most of the time, but if the drive has to do a lok of seeks, and collect db records from many areas of the disk (assuming you don't use an array or other method of speed-up) then your drive won't be bursting at full rate, since it will disconnect until the data is ready. This can then be the bottleneck, regardless of bus speed... seek times kill. Then you actually have to manipulate the data before you send it out, but that's usally an order of magnitude or three less than the disk access... drive seek: 10^-3 CPU cycle: 10^-9 (so even 100 cycles is still *way* faster than a disk access
Ah, I wasn't considering the radiation angle... I'd hope that the metal casing around the drive and the system case would be sufficient for most of this... most background radiation should be defeated by these, especially if you consider the additional metal surrounding it for mounting... three layers of metal. Not too bad.
I use my Denon D350's at work (with the RS/6k CD drive), and a set of Grado SR-120s at home... (as we go further and further off topic)... I listen to all my CDs through my stereo also (Rotel CD, amp - still waiting for my Vandersteen 2ce's). The amount of noise is a lot better with the better filtering on newer sound cards and newer motherboards (not to mention smoother supplies), which does help a lot. But again, you get used to the 'steak' pretty quickly (I just moved, and I haven't found the right position for my speakers yet - it's driving me a little insane), so computer world audio is a good step down... and you can't even blame it on road noise 8^D
Gotta say that sound cards have come a long way from my Pro-Audio Spectrum and Gravis Ultrasound... and the PC Speaker on my XT running Civilization 8^)
"Ah yes", he says while looking at his own Mat Sci book... The cold flow rate is much higher for older, non-tempered glass, this is true. However, it's not quite to the point of whole inches, it's still far less. The crystal structure of normal soda glass (SiO2) isn't nearly as intertwined as some polymer chains, and is rather layered in many cases (especially with older windows). I haven't seen many 400 year old windows lately, but the 100 year old ones certainly do not show large signs of flow. The flow constant changes with glass composition (including doping - like Pyrex), heat, stress, etc, but is several orders of magnitude below obvious change during a lifespan. If I could see some glass that is 'dripping' over a sill, then I'd be worried;-)
Yup, I want SCSI, too - and (see some other post somewhere on this story) I'm sure that there will be an Ulrastar U2W coming out soon enough (well, not soon *enough*, but soonish).
My C-64 had/s three channel sound - I thought that was pretty killer compared to the Apple ][s of the day...
The real problem (as I see it) is having the sound card running on the same power supply as the rest of the system - not always the sound card introducing noise itself. The Macs fall into the same camp here, there's still a lot of switchin noise...
I suppose the 1337 d00dz have to have the newest, best, most artifical multi-channel setup available for games (I do like 4-channel for some games, myself), but I've still got my Ensoniq Soundscape (from 1995) in my box, and it still has as good/better MIDI than the SB Live Value I have... not a lot of improvement in this area by most cards (see similar rant on 2D graphic speeds on hte Matrox G400 story).
I agree, people go a little crazy on the MP3s (high bitrate + joint stereo = why?), and I think that most people don't/can't hear (don't care about) the differences in a lot of the higher bitrate stuff. Compare 44.1/16 to 96/24, and most people can't tell much of a difference (especially on cheap equipment) for home audio stuff, and the higher encoding rates on MP3s get lost on the crappy equipment. If you are running an outboard Mark Levinson (or NAD, Rotel, Parasound, Adcom, etc) DAC hooked to a nice amp (maybe not quite a Krell or McIntosh) running to a decent set of speakers (Vandersteen, Theil, B&W, etc.), then you might hear a difference, but Joe Blow with a soundcard output driver runing into the speaker that "came with my computer" or an Aiwa bookshelf unit is going to lose all of the better information anyway, so yes, it is pointless...
NIN does have a lot of high frequency stuff (random noisy clanks and such) that don't sound quite the same in lower bitrates, but I won't argure your point on that 8^)
Mmm, lets just send all of the company's vital data across the internet from "a friend". I'll keep mine with a data vault, who has all sorts of legally binding contracts. That and a few hundred gigs is going to take far too long over DSL/cable. The point was that in most cases, you aren't going to remove the drive you are backing up to, so it isn't an adequate solution. I suppose you could use a hot swap bay, but then you would probably rather use a DLT or DAT autoloader (24GB dat * 6 isn't too bad), and the transfer rates *far* exceed that of a WAN link, assuming you want to grind your entire network to a halt to restore one server (robbing Peter to pay Paul sorta thing).
It shouldn't be long until the Ultrastar line show the 10k U2W versions of these, I would think (I have no real information). But with that 444Mb/s media transfer rate, two of these drives could saturate a U2W link... especially with the 2MB cache on it.
One issue is that for the larger servers (read: psycho RAID setups), the advantages are gained by having the most arms (physical drives), so all of the seek times are lowered, and transfer rates can be maxed out. Many people reluctantly started moving to 9GB and 18GB arms for their RAID systems - more capacity, but for heavy database usage, you want more arm for the same capacity point - the gain in performance is more important than the rise in price in many cases. You could trow together a pretty massive tower with 40s or 75s, and the transfer rates are really good, but again, the performance of the system as a whole is important. If you are building a 2TB db, would you rather have your data spread across ~60 drives (40GB) or ~265 (9GB). Data safety and performace both call for smaller individual arms here.
Well, there *is* a such thing as lossless compresion, where the original wav data is perfectly reconstructed, so there's no need to store the whole wav.
And speaking from an audiophile perspective, the quality of computer soundcards / speakers (even the *nice* ones) causes far more loss than the mp3 compression (the best codecs, anyway). If you want halfway decent sound from your computer, you need an outboard DAC - computer sound cards are all very noisy, even the SB Live Value/X-Gamer/Platinum/Dilithium (which so many seem to love). You have to get the data stream out of the box with the noisy power supply and the millions upon millions of switching transistors (and especially hard drives - you can hear most drives seeking in most sound cards without even trying too hard).
Of course, with all of the crap background noise in Quake, etc you hardly notice it, but somehow a quality recording just doesn't come across as well...
True, but off-site backups are the real issue, with father->grandfather rotations. If your office goes up in smoke, both hard drives are melted... not good. Unless of course, you intend to get three or four drives, and ship them offsite every night/week, like with tapes...
Note the debunking of the glass flow issue below, and on many urban myth pages. It may be a liquid, but the flow rate isn't visible within a lifetime. Old windows are thicker at the bottom than at the top because they made them that way (poorer methods) and installed them with the thicker ends down (for support). Considering the drive will probably have a 5 year warantee and a 7-10 lifespan, I wouldn'd worry about it too much.
yeah, I posted it yesterday morning with IBM, ZDNet and SJMercury links, along with more about the glass platters (which is hte cool part, after all)... oh well...
Glass is far more stable than the aluminum is, in terms of size / distribution. Remember that aluminum, being a metal, is far more size sensitize to heat (and it does get hot in there) that glass is. The glass surface is smoother and doesn't create as much heat spinning through the air, either, so that's reduced even more. So, throughout the lifetime of the drive, the glass should be more reliable than the old aluminum platters.
When the drive is off, the R/W heads are parked away from the platters, and the spindle is shock-absorbed, as to offset the glass breakage factor. Normally the R/W heads are parked somewhere on the platter, and vibration/impact can cause contact between the heads and the platters. This is supposed to reduce those problems rather drastically. I've been told that the glass platters can be used in laptops, but I don't kow what kind of glass it is...
7,200 rpm, six capacities: 75/60/45/30/20/15 GB, 1 to 5 glass diskplatters, 11.2 billion bits areal density, 8.5 ms average seek time, 444 Mb/s maximum media data rate, up to 100 MB/s host data rate, 2 MB buffer, 3.0 to 3.6 Bels, giant magnetoresistive (GMR) recording heads, load/unload technology.
Considering a 1.6GB solid state drive costs >$2k right now, even the larger size won't pull hte prices down that much, and the amount of storage you can get for the price in DMA (or even SCSI) relative to the solid state is rather amazing...
The new glass platters allow for greater areal density, decreasing the error rate. Check out <A HREF="http://www.ibm.com/Press/prnews.nsf/jan/2EC<nobr>0<wbr></wbr></nobr> 1AB873630970852568A3004E28F6">the press release</A> for a litle more on the platters... I'll try to find a more technical article, but the gis of it is that the glass doesn't expand like the aluminun does with heat, so your data can be closer together on the drive. There's also less friciton with the air, so there isn't as much heat to start with.
I was under the impression that the number of *people* using Linux was over 10 million, which would lead to a much larger number for the total # of linux boxen - probably a few times that, given the number of people with multiple boxes, the amount used as webservers, newsservers, etc, and the amount at buinesses... but UNIX is everywhere - as I type this on AIX from a building that amazingly has mostly AIX boxes and Thinkpads (one guess). There are a fair number of linux boxes around site...
Might have to add 2-5 for the # of S/390s running linux, too;-)
I'm glad lucent mentioned the DNS & Bind book from O'Reilly - a great reference... a standard book for what (hopefully) is a standard implementation. The exit of NetBIOS and WINS is a very Good Thing(TM), and there will be less useless broadcasts cluttering up local nets... of course, you will still have the old client machines that aren't all new and happy yet, but change will come... eventually... really...
I've heard some strange stories about the W2k DNS stuff, such as a W2k PDC having trouble if the DNS server for the 2k domain wasn't another W2k box, but I haven't experienced this personally. I don't like hte way they've integrated the DNS with Active Directory - this causes a lot of problems when upgrading an old NT4 domain (it has the most problems with underscores, which aren't valid DNS characters, but were just fine (and previously preferred) in windows domains... you can work around it, but in a large domain, it can be a lot of work getting everything back up and running happily again.
The SRV records are nothing new, but are rather useful, though many implementations don't rely on them all that much - W2k seems to take advantage of this a little more, and I think they should be applauded for that (I just said something positive about M$?!). However, they use more underscores in the SRV records, and, as the article says, many peeople are concerned about it... there used to be a forum at dnspolicy.com, but I can't seem to access it anymore...
Some good, some bad - mostly (but not all) standard, and Active Directory +DNS = yikes...
Well, if people (M$ or companies) cared about making the MCSE cert worthwhile, they would make a separate Inactive DirectXory exam, which would actually be difficult and twice as long as the others... from what I've seen, the biggest problem upgrading your NT domain to w2k is AcDir (aside from the process of it, which is ludicrous), and people should properly understand how to do any of this before they get a little paper from Bill. For now NT/W2k may be a fact of life... but rebooting isn't always the best answer...
I suppose... depending on your school and major... I had several of the MCSE exams under my belt long before I graduated from college (far easier than Network Programming, or TCP/IP design and implementation), and I didn't even bothre putting on my resume. Why? I listed my UNIX skills, and I was looking for more of a hardware/microcode job, though several large companies were throwing large sums of money at me for IT jobs, which I wasn't really after.
Real experience in the job is more useful than a cert (anybody could pass these things with a day or two of prep), and if that's really needed to get your foot in the door, do you really want that job? I suppose it's a little different when you are graduating with a computer engineering degree from a highly regarded center of higher learning than if you are out in the field trying to get a new job, but I'm convinced that unless you think the MSCE is the greatest thing since sliced bread then you shouldn't want a job where they require it... you'd be happier somewhere else.
Slashdot Mirror
Well, this comes to ~50-55 MBs xfer rate from the drive media to the cache on the drive... of course you can burst from the 2MB cache at up to bus speed (more or less), and if you are on a slower bus (plain old UW, etc) you could do a non-cached burst... that being said, if all you are doing is serving static files, then - yeah, the bottleneck isn't the drive most of the time, but if the drive has to do a lok of seeks, and collect db records from many areas of the disk (assuming you don't use an array or other method of speed-up) then your drive won't be bursting at full rate, since it will disconnect until the data is ready. This can then be the bottleneck, regardless of bus speed... seek times kill. Then you actually have to manipulate the data before you send it out, but that's usally an order of magnitude or three less than the disk access...
drive seek: 10^-3
CPU cycle: 10^-9 (so even 100 cycles is still *way* faster than a disk access
Do you have any links to IP-over-SCSI implementations? Sounds neat.
Ah, I wasn't considering the radiation angle... I'd hope that the metal casing around the drive and the system case would be sufficient for most of this... most background radiation should be defeated by these, especially if you consider the additional metal surrounding it for mounting... three layers of metal. Not too bad.
I use my Denon D350's at work (with the RS/6k CD drive), and a set of Grado SR-120s at home... (as we go further and further off topic)... I listen to all my CDs through my stereo also (Rotel CD, amp - still waiting for my Vandersteen 2ce's). The amount of noise is a lot better with the better filtering on newer sound cards and newer motherboards (not to mention smoother supplies), which does help a lot. But again, you get used to the 'steak' pretty quickly (I just moved, and I haven't found the right position for my speakers yet - it's driving me a little insane), so computer world audio is a good step down... and you can't even blame it on road noise 8^D
Gotta say that sound cards have come a long way from my Pro-Audio Spectrum and Gravis Ultrasound... and the PC Speaker on my XT running Civilization 8^)
Oh yeah - forgot about he 72ZX and the 36LZX... 10k, Ultra 160 and 160+ SCSI, FC.... mmmm
"Ah yes", he says while looking at his own Mat Sci book... The cold flow rate is much higher for older, non-tempered glass, this is true. However, it's not quite to the point of whole inches, it's still far less. The crystal structure of normal soda glass (SiO2) isn't nearly as intertwined as some polymer chains, and is rather layered in many cases (especially with older windows). I haven't seen many 400 year old windows lately, but the 100 year old ones certainly do not show large signs of flow. The flow constant changes with glass composition (including doping - like Pyrex), heat, stress, etc, but is several orders of magnitude below obvious change during a lifespan. If I could see some glass that is 'dripping' over a sill, then I'd be worried ;-)
Yup, I want SCSI, too - and (see some other post somewhere on this story) I'm sure that there will be an Ulrastar U2W coming out soon enough (well, not soon *enough*, but soonish).
My C-64 had/s three channel sound - I thought that was pretty killer compared to the Apple ][s of the day...
The real problem (as I see it) is having the sound card running on the same power supply as the rest of the system - not always the sound card introducing noise itself. The Macs fall into the same camp here, there's still a lot of switchin noise...
I suppose the 1337 d00dz have to have the newest, best, most artifical multi-channel setup available for games (I do like 4-channel for some games, myself), but I've still got my Ensoniq Soundscape (from 1995) in my box, and it still has as good/better MIDI than the SB Live Value I have... not a lot of improvement in this area by most cards (see similar rant on 2D graphic speeds on hte Matrox G400 story).
I agree, people go a little crazy on the MP3s (high bitrate + joint stereo = why?), and I think that most people don't/can't hear (don't care about) the differences in a lot of the higher bitrate stuff. Compare 44.1/16 to 96/24, and most people can't tell much of a difference (especially on cheap equipment) for home audio stuff, and the higher encoding rates on MP3s get lost on the crappy equipment. If you are running an outboard Mark Levinson (or NAD, Rotel, Parasound, Adcom, etc) DAC hooked to a nice amp (maybe not quite a Krell or McIntosh) running to a decent set of speakers (Vandersteen, Theil, B&W, etc.), then you might hear a difference, but Joe Blow with a soundcard output driver runing into the speaker that "came with my computer" or an Aiwa bookshelf unit is going to lose all of the better information anyway, so yes, it is pointless...
NIN does have a lot of high frequency stuff (random noisy clanks and such) that don't sound quite the same in lower bitrates, but I won't argure your point on that 8^)
Mmm, lets just send all of the company's vital data across the internet from "a friend". I'll keep mine with a data vault, who has all sorts of legally binding contracts. That and a few hundred gigs is going to take far too long over DSL/cable. The point was that in most cases, you aren't going to remove the drive you are backing up to, so it isn't an adequate solution. I suppose you could use a hot swap bay, but then you would probably rather use a DLT or DAT autoloader (24GB dat * 6 isn't too bad), and the transfer rates *far* exceed that of a WAN link, assuming you want to grind your entire network to a halt to restore one server (robbing Peter to pay Paul sorta thing).
Check this out for more discusion: http://slashdot.org/article.pl?sid=00/03/05/234123 2
It shouldn't be long until the Ultrastar line show the 10k U2W versions of these, I would think (I have no real information). But with that 444Mb/s media transfer rate, two of these drives could saturate a U2W link... especially with the 2MB cache on it.
One issue is that for the larger servers (read: psycho RAID setups), the advantages are gained by having the most arms (physical drives), so all of the seek times are lowered, and transfer rates can be maxed out. Many people reluctantly started moving to 9GB and 18GB arms for their RAID systems - more capacity, but for heavy database usage, you want more arm for the same capacity point - the gain in performance is more important than the rise in price in many cases. You could trow together a pretty massive tower with 40s or 75s, and the transfer rates are really good, but again, the performance of the system as a whole is important. If you are building a 2TB db, would you rather have your data spread across ~60 drives (40GB) or ~265 (9GB). Data safety and performace both call for smaller individual arms here.
Well, there *is* a such thing as lossless compresion, where the original wav data is perfectly reconstructed, so there's no need to store the whole wav.
And speaking from an audiophile perspective, the quality of computer soundcards / speakers (even the *nice* ones) causes far more loss than the mp3 compression (the best codecs, anyway). If you want halfway decent sound from your computer, you need an outboard DAC - computer sound cards are all very noisy, even the SB Live Value/X-Gamer/Platinum/Dilithium (which so many seem to love). You have to get the data stream out of the box with the noisy power supply and the millions upon millions of switching transistors (and especially hard drives - you can hear most drives seeking in most sound cards without even trying too hard).
Of course, with all of the crap background noise in Quake, etc you hardly notice it, but somehow a quality recording just doesn't come across as well...
True, but off-site backups are the real issue, with father->grandfather rotations. If your office goes up in smoke, both hard drives are melted... not good. Unless of course, you intend to get three or four drives, and ship them offsite every night/week, like with tapes...
Note the debunking of the glass flow issue below, and on many urban myth pages. It may be a liquid, but the flow rate isn't visible within a lifetime. Old windows are thicker at the bottom than at the top because they made them that way (poorer methods) and installed them with the thicker ends down (for support). Considering the drive will probably have a 5 year warantee and a 7-10 lifespan, I wouldn'd worry about it too much.
yeah, I posted it yesterday morning with IBM, ZDNet and SJMercury links, along with more about the glass platters (which is hte cool part, after all)... oh well...
hmmm, another nice job by extrans... used to work...
Glass is far more stable than the aluminum is, in terms of size / distribution. Remember that aluminum, being a metal, is far more size sensitize to heat (and it does get hot in there) that glass is. The glass surface is smoother and doesn't create as much heat spinning through the air, either, so that's reduced even more. So, throughout the lifetime of the drive, the glass should be more reliable than the old aluminum platters.
When the drive is off, the R/W heads are parked away from the platters, and the spindle is shock-absorbed, as to offset the glass breakage factor. Normally the R/W heads are parked somewhere on the platter, and vibration/impact can cause contact between the heads and the platters. This is supposed to reduce those problems rather drastically. I've been told that the glass platters can be used in laptops, but I don't kow what kind of glass it is...
The listed info says the following:
7,200 rpm, six capacities: 75/60/45/30/20/15 GB, 1 to 5 glass diskplatters, 11.2 billion bits areal density, 8.5 ms average seek time, 444 Mb/s maximum media data rate, up to 100 MB/s host data rate, 2 MB buffer, 3.0 to 3.6 Bels, giant magnetoresistive (GMR) recording heads, load/unload technology.
Considering a 1.6GB solid state drive costs >$2k right now, even the larger size won't pull hte prices down that much, and the amount of storage you can get for the price in DMA (or even SCSI) relative to the solid state is rather amazing...
The new glass platters allow for greater areal density, decreasing the error rate. Check out <A HREF="http://www.ibm.com/Press/prnews.nsf/jan/2EC<nobr>0<wbr></wbr></nobr> 1AB873630970852568A3004E28F6">the press release</A> for a litle more on the platters... I'll try to find a more technical article, but the gis of it is that the glass doesn't expand like the aluminun does with heat, so your data can be closer together on the drive. There's also less friciton with the air, so there isn't as much heat to start with.
Yup, that's it - I had read the Novell site one time, and I remembered that somewhere in the cruft of my mind... Thanks
I was under the impression that the number of *people* using Linux was over 10 million, which would lead to a much larger number for the total # of linux boxen - probably a few times that, given the number of people with multiple boxes, the amount used as webservers, newsservers, etc, and the amount at buinesses... but UNIX is everywhere - as I type this on AIX from a building that amazingly has mostly AIX boxes and Thinkpads (one guess). There are a fair number of linux boxes around site...
;-)
Might have to add 2-5 for the # of S/390s running linux, too
I'm glad lucent mentioned the DNS & Bind book from O'Reilly - a great reference... a standard book for what (hopefully) is a standard implementation. The exit of NetBIOS and WINS is a very Good Thing(TM), and there will be less useless broadcasts cluttering up local nets... of course, you will still have the old client machines that aren't all new and happy yet, but change will come... eventually... really...
I've heard some strange stories about the W2k DNS stuff, such as a W2k PDC having trouble if the DNS server for the 2k domain wasn't another W2k box, but I haven't experienced this personally. I don't like hte way they've integrated the DNS with Active Directory - this causes a lot of problems when upgrading an old NT4 domain (it has the most problems with underscores, which aren't valid DNS characters, but were just fine (and previously preferred) in windows domains... you can work around it, but in a large domain, it can be a lot of work getting everything back up and running happily again.
The SRV records are nothing new, but are rather useful, though many implementations don't rely on them all that much - W2k seems to take advantage of this a little more, and I think they should be applauded for that (I just said something positive about M$?!). However, they use more underscores in the SRV records, and, as the article says, many peeople are concerned about it... there used to be a forum at dnspolicy.com, but I can't seem to access it anymore...
Some good, some bad - mostly (but not all) standard, and Active Directory +DNS = yikes...
Sally Struthers now offers home correspondence courses in tv/vcr repair, air conditioning repair, MCSE, stamp colleting, literature...
Well, if people (M$ or companies) cared about making the MCSE cert worthwhile, they would make a separate Inactive DirectXory exam, which would actually be difficult and twice as long as the others... from what I've seen, the biggest problem upgrading your NT domain to w2k is AcDir (aside from the process of it, which is ludicrous), and people should properly understand how to do any of this before they get a little paper from Bill. For now NT/W2k may be a fact of life... but rebooting isn't always the best answer...
I suppose... depending on your school and major... I had several of the MCSE exams under my belt long before I graduated from college (far easier than Network Programming, or TCP/IP design and implementation), and I didn't even bothre putting on my resume. Why? I listed my UNIX skills, and I was looking for more of a hardware/microcode job, though several large companies were throwing large sums of money at me for IT jobs, which I wasn't really after.
Real experience in the job is more useful than a cert (anybody could pass these things with a day or two of prep), and if that's really needed to get your foot in the door, do you really want that job? I suppose it's a little different when you are graduating with a computer engineering degree from a highly regarded center of higher learning than if you are out in the field trying to get a new job, but I'm convinced that unless you think the MSCE is the greatest thing since sliced bread then you shouldn't want a job where they require it... you'd be happier somewhere else.